TREND WATCH: What’s Happening? With a turbulent 2018 behind us, forecasters have turned their attention to 2019 and beyond. While most outlets have adopted a bearish-to-cautious view of the future, Morgan Stanley is unflinchingly optimistic. The firm contends that favorable demographic trends, specifically a “youth boom” powered by a rising crop of “Gen Zers,” will push U.S. GDP growth to unforeseen heights by the end of the next decade. Once you account for demography, MS argues, the 2020s look far better than the picture painted by the official forecasters.
Our Take: We are unpersuaded. Morgan Stanley relies on framing, hand-waving, and faulty assumptions in its analysis. Upon closer examination, the full force of the demographic tailwinds referred to in the report won’t hit until the 2030s and 2040s. And even when they do arrive, their impact won’t be as positive as advertised. An honest assessment of the future—one built on realistic assumptions of working-age population, employment, and productivity growth rates—is a lot more sobering.
After a year marked by tariff wars, interest-rate hikes, slowing global growth, and stock-market selloffs, you’ve got to admire those who look past the headlines and find reasons to be long-term bullish on the U.S. economy. And not just long-term bullish, but long-term bullish for demographic reasons.
That’s what makes a new report out of Morgan Stanley, The Coming Youth Boom: When Generations Y & Z Combine, so noteworthy. Analysts who know the numbers tend to regard the stagnation in workforce growth over the next decade as pretty much written in stone.
Morgan Stanley begs to differ. The report’s authors contend that “demographic headwinds turn into tailwinds starting in the 2020s, and are long-term bullish for the U.S.” A new “youth boom,” they say, will push employment, consumption, and GDP significantly above the levels that official agencies like the Congressional Budget Office are now projecting. Moreover, Morgan Stanley accuses official forecasters of understating the impact that these “tailwinds” will have on U.S. GDP growth. Indeed, so big is this boom that Social Security and Medicare may be in much better shape than anyone realizes; their expected official insolvency dates may be postponed “perhaps by decades.”
This last claim triggered gee-whiz headlines in MarketWatch and Yahoo Finance. It also triggered a withering reply from Chuck Blahous, a seasoned expert on Social Security who has served on its Board of Actuaries. (While the Morgan Stanley study itself has not been officially released to the public, it has been fed to the financial media and widely circulated among clients and friends of MS.)
IS A “YOUTH BOOM” COMING SOON?
As demographers ourselves, we feel compelled to respond to these assertions. Is a “youth boom” coming soon? Does Morgan Stanley know something that the rest of us have overlooked? Have all the researchers publishing for years in mainstream institutions—like the U.S. Census Bureau, the U.N. Population Division, and the CBO—really gotten things very wrong?
Let’s start by summing up the gist of the mainstream projections. Basically, they say that the growth rate of the U.S. working-age population, which has been slowing dramatically over the last two decades, will reach a historic low during most of the 2020s before rising modestly and temporarily in the 2030s and 2040s. (See chart below.) The boomlet of the 2030s and 2040s will happen as the children of the relatively large Millennial Generation reach working age and as a relatively small Generation X starts retiring.
Right now, and throughout the next ten years, the opposite is happening: A relatively small late-wave Millennial and Homeland Generation is entering the workplace and a relatively large Boom Generation is retiring.
That leaves the 2020s as a uniquely grim decade if you’re looking for growth in the working-age population. At just +0.2% per year, the average growth rate in the 2020s will be barely above zero. This figure is vastly lower than working-age population growth in any earlier decade in U.S. history and is almost certainly lower than any future decade until (possibly) the 2050s. The pessimistic consequences for GDP growth follow suit. Unless you have good reason to expect a dramatic change in labor force participation rates or labor productivity, the slump in the working-age population flows through directly to a slump in expected GDP growth.
Surprisingly, the Morgan Stanley authors don’t offer any fundamental challenge to this view. They accept the Census population projections completely, agreeing that they’re “generally plausible.” And while they tweak the CBO’s GDP projections upwards (for reasons we’ll discuss shortly), the overall picture isn’t much changed from what you see in the above chart.
Rather, throughout most of the study, the authors support their dramatic “boom” thesis in what can only be called creative re-framing. They repeatedly emphasize that the U.S. demographic future looks better than that of many other developed countries (think: Italy or Japan or South Korea) on an absolute or comparative basis. They suggest this gap could be long-term bullish for the USD. Sure, that may make sense. But does this really amount to a “boom” in any ordinary sense of the word?
They also point out that “Generation Z” (born from 1997 to 2012) is a “boom” generation compared with the “Millennials” (born from 1981 to 1996) before them even though they are smaller per birth cohort. Why? Because, they say, these Gen Zers weren’t a bust generation like the Xers who followed Boomers. By thus lowering the bar—or just throwing the bar to the ground—the authors redefine a smaller Gen-Z youth generation as a “boom” force that the authors believe is “reshaping” how we should think about our economic future.
Another device the authors use a lot is bait-and-switch misdirection. While the headlines point to boom demographics “starting in the 2020s,” none of the actual demographic acceleration they point to occurs until the very end of the 2020s. Nothing significant happens until the next decade, as we have seen. And even what does occur in the 2030s and 2040s is hyperbolized out of all proportion. Even at its early-2040s peak, the growth rate in the working-age population remains far below its average since World War II. Indeed, it doesn’t even rise to the depths of the Gen-X young-adult “bust” of the early-1990s and just barely exceeds the Silent Generation young-adult “bust” of the mid-1950s.
So, let’s see, this “demographic boom in the 2020s” will still, in 2029, be generating slower growth than in 2019. And even at its 2041 peak, the growth rate will still be smaller than it was a decade ago, in 2010. Readers may be excused for regarding this as something less than a tsunami.
LOOKING AT THE NUMBERS
But it’s not all smoke and mirrors. The MS report does make small though significant upward revisions to the CBO baseline GDP projections. Exactly how large these revisions are is impossible to say because the authors provide no exact yearly estimates—only rough ranges by 5-year period. Nonetheless, they point to three alleged problems in the CBO analysis and try to provide corrections for each.
Faster Population Growth. The first problem they point to is population growth. They go along with the Census projections (again, see above), but they argue that the CBO assumes population growth numbers that are slightly below those of Census. This was news to us, since the CBO’s standard practice is to follow the latest Census projections pretty faithfully—and its latest methodological documentation lists no other source for population.
We compared the CBO’s 2018 population assumptions with the 2018 Census projection and found a near-perfect match. To be specific, we found that the two growth rates differed by only +0.01% per year through 2048 (which is much less than a rounding error.) The MS report authors found a larger gap of +0.07% per year. We’re not sure how they got this result. But even their gap is very small: It amounts to a cumulative 1.4% over twenty years. By itself, it is barely worth noticing.
Higher Labor Force Participation. The second and bigger problem they point to is labor force participation (LFP) rates. It’s one thing to know the population each year. It’s another to know what share of this population will be in the workforce. The MS authors don’t agree with how the CBO projects these rates.
The CBO’s approach is actually quite sophisticated. It divides the population into different projectable groups with different LFPs and then calculates how much the relative growth of each group will affect the next year’s total LFP. Altogether the CBO keeps track of 516 different groups: by age, gender, education, and race/ethnicity. The most important groups are age and gender (since an 80-year-old woman will obviously have a much lower LFP than a 45-year-old man). But that’s not all. The CBO also considers factors like family status, labor market tightness, and the cohort effect—that is, how the LFP of last year’s 44-year-old affects the LFP of this year’s 45-year-old.
Is all this complexity necessary? The CBO argues that, when back-tested, its model does a good job in modeling historical changes in total LFP over long time periods. Obviously, no model is perfect. But tracking the composition effect of subgroups would strike most analysts as a sound practice.
The MS authors will have none of it. Instead, they opt to simplify things drastically. They assume one unchanged LFP for the entire working-age population (age 20-64) in every future year. And they assume, for seniors (age 65+), an LFP that linearly increases over time at the same rate that this LFP has increased over the past 20 years.
Neither of these assumptions make any sense. Within each of these vast groups (over and under age 65), we can foresee with certainty large changes in age composition which can’t help but affect overall LFP. Choosing to be blind to these changes can’t possibly be an improvement to the CBO model.
The rising LFP assumption for seniors is especially perverse. It's true, as the authors point out, that the LFP of seniors has risen substantially over the last twenty years. Yet the two main drivers behind its recent rise will either stop or reverse direction over the next twenty years.
The first driver is growing economic need among 65+ Americans who (as we have shifted from Silent to Boom Generation seniors) have become more work-centric and financially vulnerable than they used to be. (See: “Boomers Stuck in ‘Low-Quality’ Jobs,” where we review the litany of reasons.) This driver is generational. It is unlikely to keep senior LFPs rising as fast over the next twenty years.
The second driver is age composition. Since the large 1946 birth-year cohort began reaching age 65 in 2011, Boomers have been tilting the center of gravity of seniors toward their younger edge. They have driven huge percentage increases in the 65-69 and 70-74 age brackets. These are the ages at which seniors in any generation are most likely to work. The older (Silent) age brackets remain much smaller by comparison. Over age 75, seniors in any generation will have negligible LFPs.
But watch what happens over the next twenty years. All those massive Boomer cohorts will move past age 75—which basically means out of the workforce. And replacing them under age 75 will be an exceptionally small generation of Gen Xers. In effect, the MS authors are expecting this diminutive Xer cadre to keeping pulling the LFP of all seniors upward, even as the center of gravity of seniors moves beyond the age at which anyone works. The authors also insist that Boomers and Xers, as they age, will improve the health status of seniors and miraculously enable a greater share of them to feel great about working. Yet they provide no evidence for this claim. Most data, to the contrary, suggest that Boomers are ushering in an era of growing health risks for 60-somethings (see: “Boomer Malaise”). We now live, after all, in an era of declining longevity.
Our take is that the age-composition effect alone will likely bring the overall LFP of seniors down over the next decade. By the mid-2030s, when even the youngest of the demographic “baby boom” are hitting their mid-70s, it is hard to imagine the same share of all seniors working as they do today. A much higher LFP would be astonishing. The MS authors talk about a “youth boom.” But for their projections to bear out, what they really need is a miraculous “senior boom” of employment.
Higher Productivity. That’s not all. The MS authors also suggest repeatedly that productivity growth will likely be higher than mainstream researchers, like those at the CBO, are forecasting. For example, they write: “We view demographic trends as a positive for U.S. equities… via their potential effects on GDP growth and labor force productivity.”
Why? Well, they argue in part that faster GDP growth generated by their proposed “demographic boom” will itself push up productivity growth. It’s a sort of positive feedback dynamic. More production and income will boost consumption and margins, and these in turn will lower the equity risk premia, pull in savings from around the world (apparently, we don’t do enough of that already!), bolster capital formation, raise the real interest rate, and avert the risk of liquidity traps.
They also argue, based on attitude surveys of “Gen Z,” that the next generation of young people entering the workforce will raise productivity because they will get along so well with “Gen Y” managers. “The shared characteristics and values of Gens Y and Z suggest they may get along better in the workplace, with positive implications for productivity.” We take great interest in generational explanations. And we tend to agree that Millennials and Homelanders are likely to get along well in the workplace. But let’s be cautious. These findings and their effects are entirely speculative. Furthermore, few Millennials will likely hold top managerial positions in large firms until the 2030s—so again we’re talking about something happening more than a decade away.
The MS authors do not attempt to quantify any of these productivity-enhancing effects. They’re basically hand-waving. Neither do they commit to any bottom-line estimate of how much the CBO is underestimating GDP growth in its projections. Yet they do imply that the CBO, on top of its “pessimistic” labor force forecast, is similarly pessimistic in its outlook on productivity. This is untrue. The CBO in fact projects that that labor productivity growth, which has declined steeply over the past 15 years, will rebound sharply in the near future. Over the past eight years, from 2009 through 2018, nonfarm labor productivity has been growing by about +1.3% annually. By the mid-2020s, the CBO expects that to climb back to about +1.8% and stay there indefinitely. That’s not pessimism.
Here’s the bottom line on the MS report’s quantitative analysis. The authors argue for a modest upward revision in the CBO baseline labor-force projection for reasons that, in our opinion, have little merit. They further urge that productivity assumptions be boosted on the basis of these modest upward demographic revisions. And they wrongly imply that the CBO has a pessimistic outlook on productivity. As we have seen, the CBO already builds a buoyantly upward reversion-to-mean into their assumptions about future productivity.
HOW TO SIMPLIFY THE LONG-TERM FORECAST
At this point, the reader may be asking: Isn’t there a simpler and more comprehensible way to look at our long-term economic future?
We think there is. Let’s imagine that our best guess of future GDP growth is the product of three growth rates. The first is the annual growth in the working-age population. The second is the annual growth in the employed share of the working-age population. And the third is the annual growth in labor productivity.
Working-age population can be forecasted for decades into the future with fairly good accuracy, because (absent immigration) everyone who will enter the workplace over the next twenty years has already been born and can be counted. Net immigration of course cannot be forecasted as easily, but Census adjusts its immigration figures regularly (though sometimes with a lag) according to changing demographic trends abroad and changing political currents at home.
Now let’s simulate a “fundamental GDP” estimate in each year by assuming that GDP growth will be equal to each year’s growth rate in this forecastable working-age population times the prior trailing 10-year average annual growth rate in the two other variables: employed share of working-age population and labor productivity. The intuition here is that, even though we cannot predict these two variables in the next year, we can get a pretty good estimate by taking the rolling trend over the last decade.
As you can see, our fundamental GDP estimate tracks actual GDP growth pretty well on a 10-year trailing average annual basis. It tracks actual “potential” GDP growth even better, because this gets rid of cyclical macro events like booms and recessions.
In the following chart, we show the same fundamental GDP growth rate, only here we decompose that rate into its three component parts: working-age population growth, working-age employment ratio, and productivity growth. There’s a lot of history reflected in these bars.
- We notice, for example, how the huge demographic boost from coming-of-age Boomers in the late 1970s (sky blue) and more working women (green) was more than offset by very disappointing “stagflation” productivity performance (navy blue).
- We notice the restraining impact that “baby bust” Gen Xers had on GDP growth in the late 1980s and early 1990s—a time when improving productivity was canceled out by weakening demographics.
- We also notice how, over the last fifteen years, our economy has been suffering from a structural triple whammy: Productivity, working-age population, and the employment-population ratio have all been moving in a negative direction.
Our model is far from perfect. Predictably, because our model uses 10-year trailing productivity performance to fix actual productivity in the next year, our fundamental GDP estimate underperforms when the actual yearly productivity rate is rising above the recent average—like in the mid-1960s or the late 1990s. And it overperforms when the actual yearly productivity rate is falling below the recent average—like in the mid-1970s. In recent years, the overperformance effect of declining productivity growth has been roughly neutralized by the underperformance effect of rising senior employment (which is outside our model).
Still, our simple model does a good job reflecting the broad contours of long-term economic growth.
Now let’s see what the model implies for the future. We get working-age population growth for each future year by consulting the latest Census projection. And we get an estimate for productivity growth and the employment-population ratio by repeating forward our last best guess: the trailing average rate for the 2009-2018 decade. The results (see next chart) are instructive. The hugely constraining impact of the very low projected growth rate in the working-age population is obvious.
Demographic stagnation, in turn, means that the lion’s share of GDP growth is going to have to come from either a higher employment rate or a speeding up of real output per worker. But, by assumption, this model keeps us stuck on our recent low 10-year average trend for both variables. It points to no change at all in the employment rate and a historically low productivity growth rate (1.3% per year). So we don’t get much help from either.
This future is not a happy one. Real GDP growth, on a 10-year rolling average basis, remains under 1.4% throughout the 2020s. It rises a bit higher in the 2030s and 2040s (peaking at about 1.7%) before falling back down again in the 2050s.
How does this compare with other mainstream estimates? Well, the IMF, in its 2018 U.S. projection, estimates that in 2023 (at the limit of its horizon) U.S. GDP growth will be 1.4%. Our estimate is 1.3%. The CBO publishes longer-range projections. Ten years out, in 2027, the CBO estimates 1.8%. Our estimate for 2027 is still 1.3%. The Fed does not publicize yearly GDP growth projections, but its “longer run” projection is 1.8%—the same as the CBO’s. The Wall Street Journal’s consensus forecast for 2021 is now also 1.8%. President Donald Trump’s OMB, meanwhile, assumes 3.0% GDP growth in every year beyond 2020.
Whether our projection is optimistic or pessimistic depends on your point of view. If you think that a good GDP projection should incorporate the best demographic forecast and choose a neutral, experience-based rule of thumb for projecting productivity (like a rolling ten-year average), then our approach seems realistic. Clearly, future productivity growth is the big unknown. Our method assumes some permanent boost over the last few years—from 1.0% in 2009 through 2018 back to 1.1% in future years. That’s a boost of +0.1%.
Compared to the IMF, we’re pretty much on the same page. The CBO, on the other hand, is more optimistic. Our productivity rate boost is +0.1%, whereas the CBO’s boost (as we saw earlier with nonfarm private output per hour) is higher—in this case about +0.4%, back up to 1.4%. The CBO also assumes a slightly higher (+0.1%) growth in the workforce. The Fed, unlike the CBO, does not publicize its projection assumptions, but we can assume they are similar to those of the CBO. The OMB says almost nothing about the assumptions underlying its very high estimate. We can only assume that OMB director Mick Mulvaney and his staff are under terrific political pressure to keep outyear GDP growth high and outyear budget deficits low.
As always, it bears repeating that a projection is not a prediction. If we can be certain of anything, it’s that the future won’t look exactly like any projection we make today. Rather, a projection should be regarded as a rational way to assess the most likely “mean path” between upside and downside accidents. The world is stochastic; it moves toward war or peace and from depression to speculative mania on a random basis. Even many of the structural drivers of long-term (or “potential”) GDP growth are not well understood.
The most uncertain driver is productivity growth. Some share of productivity growth can be linked to measurable (human or physical) capital inputs, but most seems driven by unexpected peaks and valleys in innovation that defy any measurement. Fed Chairman Jerome Powell recently declared that “total factor productivity growth is notoriously difficult to predict.” He then added for emphasis, “The record provides little basis to believe that we can accurately forecast the rate of increase in productivity.”
To the extent that OMB tries to justify its very optimistic outlook, it typically points to the possibility of a dramatic productivity revival driven by President Trump’s economic policies. No one can definitively rule out that scenario. But it does seem unwarranted to bank on a large and long-term improvement totally outside the bounds of recent experience. Assuming some upward regression toward the mean is plausible. Assuming a miraculous turnaround is imprudent—if not irresponsible.
This is actually what’s so mysterious about the new Morgan Stanley report. It points to “long-term bullish tailwinds” not by making its primary argument about productivity—what we know least about. Rather, it tries and fails to make its primary argument about demography—what we know most about. And if we have to dismiss the report’s demographic reasoning, so too must we dismiss its vague invocation of a productivity boom induced by this so-called “youth boom.” The future will no doubt deal out huge surprises. But, for now, our best-guess outlook on U.S. economic performance over the 2020s decade could probably be summed up simply: Very tough sledding ahead.